Block 4: Application: polls and the 2008 election

Week 11 · Uncertainty I · Imai §7.1.3 · about 4 min

NoteHow this works

Run each cell with Run Code, look at the output, then answer the question at the end of the block. Every cell is editable; Start Over resets it. Errors are information: read them.

For each state we average the most recent polls and treat them as a simple random sample of n = 1,000, then build a 95% confidence interval. The three columns of poll.pred (estimate, lower and upper limit) are created in the background. Does the interval contain the election result?

1. Poll estimates with confidence intervals

2. Bias and bias-corrected intervals

Conclusion: What share of the 95% intervals contain the election result, and what does correcting the bias do?
Only 59% of the intervals cover the truth: they reflect sampling noise only, not bias, house effects, undecided voters and non-random samples. Errors also vary by state, so a constant correction is not enough.